🚀 go-pugleaf

In article <6nregk$pe9$1@polo.demon.co.uk>,  <john@polo.demon.co.uk> wrote:
>In article <Evou8x.8vG@undergrad.math.uwaterloo.ca>,
>Michael Brown <m2brown@undergrad.math.uwaterloo.ca> wrote:
>>In article <aVH8IHAhElm1EwOR@thefree.net>,
>>David James Spillett  <DJSpillett@bigfoot.com> wrote:
>>>In article <899283710.14732.0.nnrp-01.c30b3753@news.demon.co.uk>, Tony
>>>Finch <fanf@demon.net> writes
>>>>You cannot use multilink with Demon. When a login happens, any
>>>>existing call for the same nodename loses routeing, because it is
>>>>redirected to the new call. If you then drop the first call, the
>>>>second call will lose routeing, because you have logged out.
>>>
>>>Am I right in quessing that you could still send packages out via the
>>>'dead' connection... Could be useful for long uploads. Not that I have
>>>two lines, like.
>>Nope. TCP is a connection-oriented transport.
>
>True, but there's no requirement for it to be a connection via a single
>route.  It's perfectly possible (although very unlikely) to have a TCP
>connection where no two IP packets follow the same route from A to B.
>
>>For each packet, there is a confirmation packet sent back.
>
>Correct, but the confirmation packet doesn't have to go back the same
>way as the original came.
>
>>The ftp client will fail.
>
>Probably, but not for the reasons stated.

When I said that - I imagined that that the first line would have a
different IP than the second. If you can tweak the first connection to
send out IP packets with a source IP for the second line, it should
work fine. In theory. :-)
-- 
Michael ``M.'' Brown
2A Computer Science - University of Waterloo

SIGSIG -- signature too long (core dumped)